One click performs the script aloud using each character's assigned voice, with emotion direction taken from the dialogue annotations.
The cue sheet and transport are built and work now. Actually speaking the lines needs a TTS provider; without one you get the cue sheet, timings and voice assignments, and the app says audio generation is not wired rather than playing silence.
This is the highest-leverage iteration tool in the studio: hearing a scene is dramatically faster and cheaper than rendering it, and most dialogue problems are audible long before they are visible.